WomenHealth AI Assistant Project

💡 Inspiration

Our project was born from a personal experience that highlighted a critical gap in healthcare information accessibility. When a friend was diagnosed with a medical condition, we observed two significant problems:

1. Difficulty Finding Authoritative Information

   • Medical websites often provide complex, technical information
   • Hard to find reliable sources among numerous unofficial websites
   • Language barriers in accessing international medical resources

2. Unreliable AI Responses

   • Existing AI chatbots often provide inconsistent medical advice
   • No guarantee of information accuracy
   • Potential risks from incorrect medical information

These observations led us to realize the crucial need for an AI agent that could provide accurate women medical information.

🛠️ How we built it

Data Collection Challenges

Our biggest initial challenge was data collection. We went through several iterations:

1. First Attempt: Traditional web scraping

   • Faced issues with website structure
   • Difficult to validate information sources
   • Inconsistent data formats

2. Final Solution: Gemini API Integration

   • Used Gemini API to generate structured Q&A pairs
   • Verified against FDA database
   • Created a comprehensive dataset of validated medical information

Model Development

After securing quality data, we:

1. Fine-tuned Gemini model using:
   • FDA-verified training data
   • Structured response formats
   • Medical accuracy checks

🎯 Challenges we ran into

1. Data Quality Assurance
   • Verifying medical accuracy of generated Q&A pairs
   • Maintaining consistency in responses
   • Balancing technical accuracy with understandability

Built With

• fda
• gemini
• python
• visual-studio